Abstract: Both geographic information system (GIS) data and remote sensing imagery are spatial data with multi-scale features although they are concerned with different types of information. This paper discusses the multi-scale segmentation of high-resolution synthetic aperture radar (SAR) images near coastal zones constrained by GIS vector data. To obtain the first level segmentation results at the county-level scale, the high-resolution SAR image is segmented using the fractal network evolution method with the constraint of GIS data. Then, at the province-level scale, the first level results are aggregated to obtain the second level segmentation results. This segmentation results satisfy both the constraint of GIS vector data and the scattering characteristics of SAR images. False coastlines in the SAR images are eliminated. Experimental results show that the proposed method is effective, and the segmentation results can be used for further image statistics and analyses.

Key words: multi-scale, segmentation, coastal zone, remote sensing, geographic information system (GIS), synthetic aperture radar (SAR)